Jeffrey Kagle

Rigid motors and hydraulics struggle under deep-water pressure and tend to be bulky and inefficient, while conventional dielectric elastomer actuators deliver high force but little displacement. We designed a deformable rotary motor driven by liquid dielectric elastomer actuators (Peano-HASELs), selected the concept through weighted trade studies, and built custom quarter-circle actuators and a 3D-printed motor housing as a proof of concept.

Key results

1. Weighted trade studies selected liquid DEAs as the actuation method (score 2.9, ahead of solid DEAs at 2.7, motors at 2.6, and hydraulics at 2.0) and rotary motion as the movement method (3.1 vs. 2.45 for flapping and 2.2 for jetting).
2. Designed a three-pouch, quarter-circle Peano-HASEL actuator with painted carbon electrodes so four units fit around a circular frame and pull on an eccentric central rotor.
3. Developed the fabrication process: heat-sealing BOPP film with a 3D printer, painting carbon electrodes, injecting silicone oil as the liquid dielectric, and resealing the pouches.
4. Fabricated actuators deformed at 5, 8, and 9 kV from an off-board high-voltage DC/DC converter without damage; testing was limited to verifying functionality.
5. Identified the main manufacturing fixes for future builds: thicker seal lines in the printer path to prevent leaks, better-placed injection points, and fewer air bubbles in the pouches.

Figures

Snap-through buckling can deliver a fast, distinct indentation for wearable haptic feedback, but the beam needs an actuator that supplies 2.9 N over a 3.36 mm stroke to trigger the snap. With LZ Corrales, I designed a compact Lorentz-force solenoid with a permanent-magnet plunger, sized it with a magnetic-vector-potential force model, and integrated it with the beam in a 3D-printed housing. The integrated prototype delivers the impulse force and displacement needed to trigger snap-through, and the beam resets passively.

Key results

1. An off-the-shelf solenoid (12 V, 0.25 A, 0.5 N) could not displace the beam, so we designed a custom solenoid with an N52 permanent-magnet plunger to raise force density without adding mass.
2. Swept current, magnet diameter, turns per layer, and layer count in a magnetic-vector-potential force model; the selected design uses 0.5 A, a 10 mm magnet, and 120 turns × 18 layers of 30 AWG wire (85.22 m).
3. Validated the model against a commercial solenoid on an ATI Nano force sensor: modeled and measured force maxima and minima closely aligned, with discrepancies only near the coil edge.
4. The custom solenoid stays above the 2.9 N threshold over ≈19 mm of stroke when only 3.36 mm is needed, leaving room to shrink it by matching the beam's force–displacement profile.
5. Housing iterations showed that placing the solenoid and beam in series, rather than in parallel, removes a bending moment that cut the force reaching the beam; 0.1 mm clearances made the 3D-printed parts fit reliably.

Figures

Haptic cues are increasingly used in gait rehabilitation, but few studies compare feedback types head-to-head on both walking performance and user experience. With Daniel Garcia and EV Morisseau, I built The Gait Coach, a wearable that delivers vibrotactile, skin-stretch, poke, or auditory cues at 120 steps per minute while an IMU on the foot measures how quickly the wearer locks onto that pace. In a three-participant study, vibrotactile feedback came closest to the auditory control, while audio remained the most comfortable.

Key results

1. Vibrotactile cues matched the audio control most closely: Welch's t-tests on time-to-match gave p = 0.604 for vibrotactile versus audio, compared with p = 0.1730 for skin stretch.
2. Participants rated audio highest for comfort, perceptibility, and self-scored matching, with vibrotactile a close second; skin stretch ranked lowest and, with poke, was called the most confusing.
3. Built the wearable around a Hapkit board: an ERM motor, a solenoid poke, a DC-motor skin-stretch end effector, and headphones, with custom 3D-printed housings for each actuator.
4. Wrote a heel-strike detector for the foot IMU (acceleration-magnitude threshold with a refractory period) that times how long a walker takes to hold the target cadence for 6 consecutive seconds.
5. Poke testing was cut short when the solenoid's power draw disrupted the microcontroller, leaving one data point; a separate 12 V supply, sturdier connections, and better data logging are the identified fixes.

Figures

Cuttlefish swim by passing a sinusoidal wave along muscular-hydrostat fins, but no experimental study had measured or modeled the thrust these fins produce for forward swimming. This theory and literature-review project synthesized biological fin mechanics (anatomy, EMG activation, and force–strain models) with a robotic sinusoidal-fin model into one hybrid force model, creating a benchmark for comparing cuttlefish and designing soft biomimetic robots.

Key contributions

1. Built a hybrid thrust model by equating the robotic sinusoidal-fin equations of Su et al. with the biological fin-force model of Johnsen and Kier, filling the gap left by the lack of measured fin thrust data for cuttlefish.
2. Linked fin physiology to propulsion: muscles fire in a sinusoidal activation wave, gentle beating uses only the aerobic transverse muscle, and vigorous beating recruits all three muscle groups for higher force.
3. Included elastic energy stored and recovered by the muscular-hydrostat fin, and estimated energy output from work (E = F·d) since no intake-versus-output data exist for fin propulsion alone.
4. Proposed validation experiments: a soft biomimetic cuttlefish robot in a flow tank with force transducers, PIV, and strain gauges, and a live-animal flow-tank test to measure fin thrust using Newton's second law.

The hybrid thrust model

No study has measured the thrust of a live cuttlefish fin, so the model borrows it from robotics. Assuming a robotic sinusoidal fin (Su et al., 2022) with the same size and shape as the cuttlefish fin, the robot's thrust and drag equations stand in for the animal's, while the biological model (Johnsen & Kier, 1993) supplies the out-of-plane fin forces.

1. Net force from motion. Over a short time segment Δt, the change in swimming velocity u gives the acceleration, and Newton's second law gives thrust minus drag, where m is the cuttlefish's wet mass:

2. Drag at a point velocity. From the robotic model, drag is the viscous shear (fluid viscosity μ) integrated over the fin surface S:

3. Point thrust. Combining the two isolates thrust, which must also equal the pressure p the beating fin exerts on the water in the swimming direction:

4. Fin forces from the animal. The biological model gives the dorsoventral and lateromedial fin forces from tissue tension: T_c is the tension in the crossed connective-tissue fibres at angle θ, T_dv the dorsoventral muscle tension, A_tm and s_tm the transverse-muscle area and stress, and each element follows T = EAε:

5. Combined. Together these give a three-dimensional fin force in the posteroanterior, lateromedial, and dorsoventral directions:

Figures

Students struggle to stay focused through long lectures, and neither productivity apps nor current smart glasses solve it. Our team, STEN Labs, designed Focus Glasses: ordinary-looking eyewear with low-intensity 40 Hz LED flicker built into the temple arms, based on research linking 40 Hz stimulation to sustained attention. The specification covers user research, TRIZ-driven design, thermal and structural checks, conjoint-based pricing, and a five-year market forecast.

Key results

1. Grounded the concept in 50 immersion, observation, and interview sessions with Cornell students; an AHP ranking put academic achievement and focus as the most critical problem areas.
2. Designed the electronics around RGB LEDs driven by an ATtiny202 on a custom PCB in the temple arm, powered by a 3 V CR1220 cell, with TRIZ principles guiding LED placement, diffusion, and thermal management.
3. Thermal check: each LED dissipates about 22 mW, raising temperature only 3.3–5.5 °C above ambient, which is not a safety concern for skin contact.
4. Structural check: a 10 N press on the ~50 mm² temple produces about 200 kPa, orders of magnitude below the ~40 MPa yield strength of ABS.
5. Business case: value-based pricing of $59–$79 against a ~$50 early unit cost with injection-molded frames, and a Bass-model forecast over a 500,000-user market giving an NPV of $15,543,276.

Figures

A semester-long design of the TR315-r, a 1502 cc inline-3 engine for cruiser motorcycles, by a 10-person team. As project manager I coordinated the schedule and deliverables, focused on the fluid mechanics and thermodynamics behind engine cooling, and contributed to every section. The final specification covers the engine design, noise and emissions compliance, manufacturing, and a three-year business case.

Key results

1. Cooling: chose water cooling after finding that air cooling would need an excessive amount of fin area; water cooling takes less space and moves more heat. Set a 120 °C maximum engine surface temperature, well below the ~200 °C at which ABS shrouding begins to melt.
2. Noise: computed exhaust sound intensity from exhaust pressure, density, and speed of sound, then applied the inverse-square law. With a muffler (~35 dB reduction) the engine reaches 72.93 dB at 5000 rpm and 77.83 dB at redline, measured 50 ft away — within Washington's 80 dBA limit, the strictest in the US.
3. Emissions: initial output needed a 97% reduction at cruise; a 400-cell metallic catalytic converter brings HC + NOx below 0.8 g/km and CO below 6.4 g/km, meeting CARB and EPA limits.
4. Final layout: over-square inline-3 (88.9 mm bore, 76.2 mm stroke, ratio 1.1667) with a pent-roof head, port injection, and wet-sump lubrication — 280.87 lb in a 20.32 × 14.50 × 12.75 in envelope.
5. Business case: $3,277.17 production cost per engine at 15,000 units a year; a $4,099.95 sale price gives a 23.4% gross margin and 14.9% ROI over three years on a $17.55 million initial investment.

Figures

Competitive sailing hardware with a three-layer elastomeric architecture — polyurethane body, nylon rope core, and woven polymer mesh sleeve. Patent pending.